May_EDFA_Digital

edfas.org ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 2 8 and BGA bumps are clearly visible. However, because the sample is embedded in the epoxy puck severe charging artifacts are visible, resulting in a distorted image. The charging artifacts create further image distortion seen in the higher magnification image of the laminate-C4-die interface in Fig. 4c. The laminate layers appear to be non-planar and all metal features appear to have a thick dark outline. While the BEOL metal stack is clearly visible in the high magnification image in Fig. 4d, the charging has caused the appearance of horizontal lines and lateral shifts in the image as it was collected. The image distor- tion due to charging in the C4 and underfill also makes it difficult to identify defects in these regions. Figure 5 shows the corresponding SEM images for a cross-section of the sample in Config1. As in the case of Config0 in Fig. 4, severe image distortion is caused by charging artifacts. However, the polished surface quality appears to be comparable betweenConfig0 andConfig1. Config1 therefore results in 84% reduction in polishing time for a comparable surface polish quality. REVISION OF THE MODIFIED PUCK TO REDUCE ARTIFACTS The charging-related SEM image dis- tortion was observed equally in Config0 and Config1 in spite of the preliminary Cr coating, indicating that themodified puck in Config1 did not alleviate the charging problem. Because the modified puck was custom-designed, it was first proposed that metal pillars could be incorporated into the epoxy, which would provide grounding through the body of the puck. However, regardless of their dimension, distribution, or spacing, themere presence of metal pillars would cause significant wear on the sandpaper andpolishingpads. The optimal casewould have a conductive epoxy-basedmethod of grounding. Mixing conductive filler into the epoxy caused the puck to become opaque, making it diffi- cult to visually gauge polishing progress. The final solution was to make pillars and sections, rather than the entire puck, of conductive epoxy. These cured sections of conductive epoxy could be fabricated in advance and added into the final puck along with the sample. Figure 6 shows results of incorporating conductive epoxy into a modified puck with an impregnated sample, referred to as Config2. Figure 6a is the optical image of modified puck with the sample. The darker contrast of the epoxy on the left Fig. 4 SEM images of the Config0 cross-section at the center of the first rowof C4 bumps. Lowmagnification images of the (a) leftand (b) right corners of the die, respectively. (c) The laminate-C4-die interface. (d) Highmagnification image of the die-C4 interface. All images show severe distortion due to charging. Fig. 5 SEM images of the Config1 cross-section at the center of the first rowof C4 bumps. Lowmagnification images of the (a) leftand (b) right corners of the die, respectively. (c) The laminate-C4-die interface. (d) Highmagnification image of the die-C4 interface. All images show severe distortion due to charging. (continued on page 10)